Check whether the outcome improved
Verify equal arrivals
Release equivalent work evenly for every class and compare ready, claim, completion, and failure records. Confirm no class loses all service and that tie behavior is documented. Use the same worker pool and queue topology as the proposed production path.
Verify skewed arrivals
Increase ready volume for one class while retaining eligible work for the others. Check the stated fairness rule, oldest wait, and class service share. Confirm the noisy class does not consume every slot unless that is the explicit policy and other classes have a recovery state.
Verify a changing mix
Introduce a new class arrival during a run and follow service before and after it appears. Confirm class assignment and allocation remain visible, and that a late class cannot silently displace existing work. Reconcile terminal results for all classes after the interval ends.
What to carry forward
Accept when equal, skewed, and changing workloads produce the documented per-class service and wait behavior with no unexplained disappearance. Keep policy-approved allocation differences visible. Reopen verification if a new class changes service without a stated rule or if aggregate throughput hides an oldest-wait breach.
Technical background: Rails guides.
Keep the decision with the work.
Use a Work Item in Aglet to record the problem, the evidence you have, and the next decision. Add an owner and priority, then keep updates in the discussion so the next person can follow the reasoning.
Create an account See the product workflow